A business risk early warning method and related device

Abstract

This invention discloses a method and related apparatus for enterprise risk early warning, relating to the field of enterprise risk management technology. The method includes: constructing an enterprise risk relationship graph with enterprises and related entities as nodes and the relationships between nodes as edges; acquiring risk characteristic data corresponding to each node in the enterprise risk relationship graph; determining the node risk potential value corresponding to each node based on the risk characteristic data; generating a directional risk propagation path in the enterprise risk relationship graph using the node risk potential value as a guiding signal; and generating an enterprise risk early warning result based on the risk propagation path. This invention places enterprises in a relationship network, captures risk associations, and generates risk propagation paths guided by node risk potential values, revealing the direction and trajectory of risk transmission, thereby issuing early warnings before risks spread to enterprises, significantly improving the timeliness and foresight of early warnings.

Description

Technical Field

[0001] This application relates to the field of enterprise risk management technology, and in particular to an enterprise risk early warning method and related apparatus. Background Technology

[0002] Enterprise risk early warning is a crucial link in ensuring the stable operation of the economic system. Current mainstream enterprise risk early warning solutions mainly rely on static indicators or rule-based models, and their analysis objects are often limited to a single enterprise entity, conducting independent risk assessments by examining its financial indicators or operational behavior.

[0003] However, in complex business environments, enterprises form numerous interactive relationships with various related entities. Current enterprise risk warning schemes treat enterprises as isolated individuals, and their analytical perspective is therefore limited by the enterprise's own financial indicators or operating conditions. This not only leads to serious lag and limitations in warning results, but also makes it difficult to capture systemic risks caused by related networks. Summary of the Invention

[0004] In view of this, this application provides a method and related apparatus for enterprise risk early warning, which addresses the serious lag and limitations of existing enterprise risk early warning schemes, and makes it even more difficult to capture systemic risks caused by interconnected networks. The technical solution is as follows: The first aspect of this application provides a method for enterprise risk early warning, including: A risk relationship graph is constructed using enterprises and their associated entities as nodes, and the relationships between nodes as edges. Obtain the risk characteristic data corresponding to each node in the enterprise risk relationship diagram; Based on the risk characteristic data corresponding to each node in the enterprise risk relationship diagram, the node risk potential value corresponding to each node in the enterprise risk relationship diagram is determined. The node risk potential value represents the degree of risk aggregation or the potential intensity of risk propagation of the corresponding node. The starting node of the path is determined based on the node risk potential value corresponding to each node in the enterprise risk relationship diagram. Starting from the starting node, the node risk potential value is used as the guiding signal for path extension. The path is extended along the relationship between nodes to generate a directional risk propagation path. Based on the risk propagation path described, generate enterprise risk warning results.

[0005] In one possible implementation, the enterprise risk warning method further includes: Based on the risk propagation path and the enterprise risk warning results, determine the parameters of the risk response strategy.

[0006] In one possible implementation, the step of obtaining the risk characteristic data corresponding to each node in the enterprise risk relationship graph is as follows: For each node in the enterprise risk relationship diagram, obtain the multi-source risk feature data corresponding to the node; The multi-source risk characteristic data includes some or all of the following characteristic data: enterprise operation characteristic data reflecting the enterprise operation status of the corresponding node, interaction characteristic data reflecting the interaction between the corresponding node and related nodes, historical risk characteristic data reflecting the historical risk situation of the corresponding node, and structural characteristic data reflecting the position structure of the corresponding node in the enterprise risk relationship diagram.

[0007] In one possible implementation, determining the node risk potential value corresponding to each node in the enterprise risk relationship diagram based on the risk characteristic data corresponding to each node in the enterprise risk relationship diagram includes: The large model is invoked, and the node risk potential value corresponding to each node in the enterprise risk relationship diagram is determined based on the risk characteristic data corresponding to each node in the enterprise risk relationship diagram.

[0008] In one possible implementation, adjacent nodes in the enterprise risk relationship graph are configured with relationship weights; The process of extending the path from the starting node, using the node's risk potential value as a guiding signal, and extending the path along the relationships between nodes to generate a directional risk propagation path includes: Taking the starting node of the path as the current node, calculate the risk potential energy difference between the current node and each adjacent node. Based on the calculated risk potential energy difference and the relationship weights between the current node and each neighboring node, calculate the transition probability distribution from the current node to each neighboring node. Based on the calculated transition probability distribution, the next path node is determined from each of the current node's neighboring nodes to extend the path; If the preset path extension termination condition is not met, the next path node is taken as the new current node, and the calculation of the risk potential energy difference between the current node and each adjacent node and subsequent steps are performed until the path extension termination condition is met, and a complete risk propagation path is obtained.

[0009] In one possible implementation, generating an enterprise risk warning result based on the risk propagation path includes: Based on the risk propagation path, node-level risk feature sequences, path-level risk feature sequences, and theme-level risk feature sequences are constructed to obtain the hierarchical risk features of the risk propagation path. Based on the hierarchical risk characteristics of the risk propagation path, enterprise risk warning results are generated.

[0010] In one possible implementation, the step of constructing node-level risk feature sequences, path-level risk feature sequences, and topic-level risk feature sequences based on the risk propagation path to obtain hierarchical risk features of the risk propagation path includes: Based on the order of the nodes on the risk propagation path, the nodes on the risk propagation path are arranged into a path node sequence. Based on the order of the nodes on the risk propagation path, the node types of the nodes on the risk propagation path are arranged into a node type sequence. Based on the path node sequence and the node type sequence, a node-level risk feature sequence, a path-level risk feature sequence, and a topic-level risk feature sequence are constructed to obtain the hierarchical risk features of the risk propagation path.

[0011] In one possible implementation, adjacent nodes in the enterprise risk relationship graph are configured with relationship types; The step of constructing node-level risk feature sequences, path-level risk feature sequences, and theme-level risk feature sequences based on the path node sequence and the node type sequence includes: Based on the node risk potential value corresponding to each node in the path node sequence, a node-level risk feature sequence is constructed. Traverse the adjacent node type pairs in the node type sequence, form a triplet by combining the adjacent node type pairs and the relationship type between the two nodes corresponding to the adjacent node type pairs, and form a path-level structural feature sequence by combining the obtained triplets according to the order of each node on the risk propagation path. The node-level risk feature sequence is aligned and fused with the path-level structural feature sequence to obtain a path-level risk feature sequence. Each risk feature in the path-level risk feature sequence includes a pair of adjacent node types, as well as the relationship type and node risk feature of the two nodes corresponding to the pair of adjacent node types. Risk features with the same relationship type in the path-level risk feature sequence are merged to obtain the topic-level risk feature sequence.

[0012] In one possible implementation, constructing a node-level risk feature sequence based on the node risk potential value corresponding to each node in the path node sequence includes: Based on the order of the nodes in the path node sequence, the node risk potential values ​​corresponding to each node in the path node sequence are used to form a node-level initial risk feature sequence. Based on the order of the nodes in the path node sequence, a position index is assigned to each node in the path node sequence to obtain a position index sequence; The node-level initial risk feature sequence is combined with the location index and node risk potential value of the corresponding node in the location index sequence to obtain the final node-level risk feature sequence.

[0013] In one possible implementation, the step of constructing a node-level risk feature sequence, a path-level risk feature sequence, and a topic-level risk feature sequence based on the path node sequence and the node type sequence further includes: The node-level risk feature sequence is input into a node-level attention network to obtain a node-level attention weight vector; the path-level risk feature sequence is input into a path-level attention network to obtain a path-level attention weight vector; the topic-level risk feature sequence is input into a topic-level attention network to obtain a topic-level attention weight vector. The node-level risk feature sequence is weighted using the node-level attention weight vector; the path-level risk feature sequence is weighted using the path-level attention weight vector; the topic-level risk feature sequence is weighted using the topic-level attention weight vector. The weighted node-level risk feature sequence is used as the final node-level risk feature sequence, the weighted path-level risk feature sequence is used as the final path-level risk feature sequence, and the weighted theme-level risk feature sequence is used as the final theme-level risk feature sequence.

[0014] In one possible implementation, generating enterprise risk warning results based on the hierarchical risk characteristics of the risk propagation path includes: By fusing the node-level risk feature sequence, path-level risk feature sequence, and theme-level risk feature sequence in the hierarchical risk features of the risk propagation path, a comprehensive risk feature of the risk propagation path is obtained. Based on the comprehensive risk characteristics of the aforementioned risk propagation path, determine the enterprise risk score; From a set of multiple preset risk scoring intervals, determine the risk scoring interval to which the enterprise's risk score belongs; Generate enterprise risk warning results that include the enterprise risk score and the risk score range to which the enterprise risk score belongs.

[0015] In one possible implementation, determining the risk response strategy parameters based on the risk propagation path and the enterprise risk warning result includes: Construct a path parameter sequence corresponding to the risk propagation path. The path parameter sequence includes the node risk potential value corresponding to each node on the risk propagation path and the relationship weight between adjacent nodes on the risk propagation path. Based on the path parameter sequence and the enterprise risk warning results, determine the risk response strategy parameters.

[0016] In one possible implementation, determining the risk response strategy parameters based on the path parameter sequence and the enterprise risk warning result includes: Based on the enterprise risk score and risk score range in the enterprise risk warning results, determine the risk handling strategy type parameter used to indicate the risk handling method; And / or, based on the path length of the risk propagation path and the enterprise risk score in the enterprise risk warning result, determine the risk response priority parameter; And / or, determine the risk propagation blocking parameters based on the node risk potential values ​​in the path parameter sequence; And / or, determine the risk control resource allocation parameters based on the relation weights in the path parameter sequence; And / or, determine the risk monitoring update parameters based on the risk score range in the enterprise risk warning results.

[0017] A second aspect of this application provides an electronic device, including at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is used to execute the computer program so that the electronic device can implement any of the above-mentioned enterprise risk warning methods.

[0018] A third aspect of this application provides a computer storage medium carrying one or more computer programs, which, when executed by an electronic device, enable the electronic device to implement any of the aforementioned enterprise risk warning methods.

[0019] A fourth aspect of this application provides a computer program product including computer-readable instructions that, when executed on an electronic device, cause the electronic device to implement any of the aforementioned enterprise risk warning methods.

[0020] Using the aforementioned technical solution, the enterprise risk early warning method provided in this application first constructs an enterprise risk relationship diagram, then obtains the risk characteristic data corresponding to each node in the enterprise risk relationship diagram, and determines the node risk potential value corresponding to each node in the enterprise risk relationship diagram accordingly. Then, using the node risk potential value as a guiding signal in the enterprise risk relationship diagram, a directional risk propagation path is generated. Finally, an enterprise risk early warning result is generated based on the risk propagation path. The enterprise risk early warning method provided in this application considers enterprises within a relationship network, breaking through the existing analytical model limited to the situation of a single enterprise. It can effectively capture the risk correlation between enterprises. By generating a risk propagation path guided by the node risk potential value, it can reveal the direction and trajectory of risk transmission from the source outwards. Based on the risk propagation path, an early warning can be issued before the risk spreads to the enterprise, realizing a shift from post-event handling to pre-event prevention, significantly improving the timeliness and foresight of the early warning. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0022] Figure 1 A flowchart illustrating the enterprise risk warning method provided in this application embodiment; Figure 2 This is a flowchart illustrating how a directional risk propagation path is generated by starting from the path's initial node, using the node's risk potential value as a guiding signal for path extension, and extending the path along the relationships between nodes, as provided in this embodiment of the application. Figure 3 A flowchart illustrating the process of generating enterprise risk warning results based on risk propagation paths, provided for embodiments of this application; Figure 4 A schematic diagram illustrating how the weight vectors corresponding to the node-level risk features, path-level risk feature sequences, and topic-level risk feature sequences provided in this application embodiment are determined; Figure 5 A flowchart illustrating the process of determining risk response strategy parameters based on risk propagation paths and enterprise risk warning results, provided for an embodiment of this application; Figure 6 This is a schematic diagram of the structure of the enterprise risk warning device provided in the embodiments of this application. Detailed Implementation

[0023] The embodiments of this application are described below with reference to the accompanying drawings. The terminology used in the implementation section of this application is for explaining specific embodiments only and is not intended to limit the scope of this application.

[0024] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.

[0025] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not explicitly listed or inherent to those processes, methods, products, or apparatuses.

[0026] In complex business environments, enterprises form numerous interactive relationships with various related entities. However, the analytical scope of current enterprise risk warning solutions is limited by the enterprise's own operating conditions, resulting in serious lag and limitations in warning results, and making it even more difficult to capture systemic risks caused by related networks.

[0027] Given the numerous shortcomings of current enterprise risk warning schemes, the inventors of this application conducted research and, through continuous investigation, proposed an enterprise risk warning method that overcomes the deficiencies of existing enterprise risk warning schemes. The following embodiments will further illustrate the enterprise risk warning method provided in this application.

[0028] Please see Figure 1 The diagram illustrates a flowchart of an enterprise risk warning method provided in an embodiment of this application. This enterprise risk warning method may include: Step S101: Construct a risk relationship graph for enterprises, using enterprises and their related entities as nodes and the relationships between nodes as edges.

[0029] Constructing an enterprise risk relationship diagram essentially involves abstracting the intricate relationships between enterprises in the real world into an analyzable and processable network structure. Constructing an enterprise risk relationship diagram is the foundation for enterprise risk early warning.

[0030] The nodes in the enterprise risk relationship diagram not only include the enterprise itself, but also its related entities. Related entities can be capital-level entities, such as controlling shareholders, actual controllers, investment institutions, and other shareholders. Related entities can also be business-level entities, such as upstream suppliers, downstream customers, and important partners. Related entities can also be external credit enhancement entities, such as guarantors or guaranteed parties. Including these entities as nodes in the diagram aims to restore the enterprise to its true business ecosystem and avoid viewing a single enterprise in isolation.

[0031] In the enterprise risk relationship graph, each edge represents a certain relationship between the two nodes it connects. Each edge in the enterprise risk relationship graph is associated with a relationship type, which is the relationship type configured for the two nodes connected by the edge (e.g., equity relationship / supply chain relationship / guarantee relationship, etc.). Each edge in the enterprise risk relationship graph is also associated with a relationship weight, which is the relationship weight between the two nodes connected by the edge. The relationship weight between the two nodes represents the importance of the relationship between the two nodes.

[0032] Step S102: Obtain the risk characteristic data corresponding to each node in the enterprise risk relationship diagram.

[0033] Among them, risk characteristic data refers to data that can reflect the risk status of the corresponding node.

[0034] Preferably, for each node in the enterprise risk relationship diagram, the multi-source risk feature data corresponding to that node is obtained, so as to obtain the multi-source risk feature data corresponding to each node in the enterprise risk relationship diagram.

[0035] It should be noted that the multi-source risk characteristic data corresponding to any node is data that can reflect the risk status of that node in multiple dimensions.

[0036] The multi-source risk characteristic data corresponding to any node may include, but is not limited to, some or all of the following characteristic data: enterprise operation characteristic data, interaction characteristic data, historical risk characteristic data, and structural characteristic data.

[0037] Among them, the enterprise operation characteristic data corresponding to any node is the data reflecting the enterprise's operating status at that node, such as the statistical values ​​of operating fluctuations, operating scale, and operating stability. The interaction characteristic data corresponding to any node is the data reflecting the interaction between that node and related nodes, such as the statistical values ​​of interaction frequency, transaction scale, and cooperation duration. The historical risk characteristic data corresponding to any node is the data reflecting the historical risk situation of that node, such as the number of historical abnormal records, the frequency of risk events, and the statistical values ​​of risk handling records. The relational structure characteristic data corresponding to any node is the data reflecting the positional structure of that node in the enterprise risk relational graph, such as the number of related nodes of that node in the enterprise risk relational graph, the number of edges connected to that node, and the number of relational types associated with the edges connected to that node.

[0038] Step S103: Based on the risk characteristic data corresponding to each node in the enterprise risk relationship diagram, determine the node risk potential value corresponding to each node in the enterprise risk relationship diagram.

[0039] In the enterprise risk relationship diagram, the node risk potential value corresponding to any node represents the degree of risk aggregation or the potential strength of risk propagation of that node. It should be noted that the degree of risk aggregation of a node refers to the level of risk that the node itself bears or accumulates, while the potential strength of risk propagation of a node refers to the strength of the node's ability to transmit risk outward in the enterprise risk relationship diagram.

[0040] Optionally, a large model can be invoked to determine the node risk potential value corresponding to each node in the enterprise risk relationship diagram based on the risk characteristic data corresponding to each node.

[0041] Step S104: Determine the starting node of the path based on the node risk potential value corresponding to each node in the enterprise risk relationship diagram. Starting from the starting node, use the node risk potential value as the guiding signal for path extension, and extend the path along the relationship between nodes to generate a directional risk propagation path.

[0042] In one possible implementation, a risk potential energy threshold can be preset. Then, based on the preset risk potential energy threshold and the node risk potential energy values ​​corresponding to each node in the enterprise risk relationship diagram, the starting node of the path is determined from the enterprise risk relationship diagram. Specifically, for each node in the enterprise risk relationship diagram, it is determined whether the node risk potential energy value corresponding to the node is greater than the preset risk potential energy threshold. If the node risk potential energy value corresponding to the node is greater than the preset risk potential energy threshold, then the node is determined to be the starting node of the path.

[0043] After determining the starting node of the path, in the enterprise risk relationship diagram, starting from the starting node, the risk potential value of the node is used as a guiding signal for the path extension. The path is extended along the relationship between the nodes to generate a directional risk propagation path.

[0044] Step S105: Generate enterprise risk warning results based on the risk propagation path.

[0045] A risk propagation path is a directional sequence of nodes, such as enterprise node A → related entity node B → enterprise node C → related entity node D → enterprise node E. The risk propagation path contains various information, such as where the risk originates, which intermediate links it passes through, who it might affect, and the complete chain of risk transmission.

[0046] The enterprise risk early warning method provided in this application first constructs an enterprise risk relationship diagram, then obtains the risk characteristic data corresponding to each node in the enterprise risk relationship diagram, and determines the node risk potential value corresponding to each node in the enterprise risk relationship diagram accordingly. Next, using the node risk potential value as a guiding signal in the enterprise risk relationship diagram, a directional risk propagation path is generated. Finally, an enterprise risk early warning result is generated based on the risk propagation path. The enterprise risk early warning method provided in this application considers enterprises within an enterprise relationship network, breaking through the existing method's analysis mode limited to the situation of a single enterprise. It can effectively capture the risk correlation between enterprises. By generating a risk propagation path guided by the node risk potential value, it can reveal the direction and trajectory of risk transmission from the source outwards. Based on the risk propagation path, an early warning can be issued before the risk spreads to the enterprise, realizing a shift from post-event handling to pre-event prevention, significantly improving the timeliness and foresight of the early warning.

[0047] In some embodiments of this application, the implementation process of step S104 above, which involves "starting from the path starting node, using the node risk potential energy value as the guiding signal for path extension, extending the path along the relationship between nodes, and generating a directional risk propagation path", is described.

[0048] like Figure 2 As shown, the process of generating a directional risk propagation path, starting from the path's initial node and using the node's risk potential value as a guiding signal for path extension, and extending the path along the relationships between nodes, can include: Step S201: Take the starting node of the path as the current node and calculate the risk potential energy difference between the current node and each adjacent node.

[0049] For example, if the starting node of the path is A, and node A is taken as the current node, and the adjacent nodes of the current node A are B, C, and D, then calculate the difference between the risk potential value corresponding to node A and the risk potential value corresponding to node B, calculate the difference between the risk potential value corresponding to node A and the risk potential value corresponding to node C, and calculate the difference between the risk potential value corresponding to node A and the risk potential value corresponding to node D. Finally, obtain the risk potential value difference between the current node A and the adjacent node B, the risk potential value difference between the current node A and the adjacent node C, and the risk potential value difference between the current node A and the adjacent node D.

[0050] Step S202: Calculate the transition probability distribution from the current node to each neighboring node based on the risk potential energy difference between the current node and each neighboring node and the relationship weight between the current node and each neighboring node.

[0051] The probability distribution of the transition from the current node to each neighboring node refers to the probability of transitioning from the current node to each neighboring node, that is, the likelihood of extending from the current node to each neighboring node. It reflects the tendency of risk to propagate from the current node in all directions.

[0052] Step S203: Based on the transition probability distribution from the current node to each neighboring node, determine the next path node from each neighboring node of the current node to extend the path.

[0053] After calculating the transition probability from the current node to each adjacent node, based on the calculated probability, a preset selection strategy (such as selecting the adjacent node corresponding to the maximum transition probability) is adopted to select a node from each of the current node's adjacent nodes as the next stop on the path, thereby extending the current risk propagation path forward by one step.

[0054] For example, the current node (the starting node of the path) is node A. In the enterprise risk relationship graph, the neighboring nodes of the current node A are B, C, and D. The transition probability from the current node A to the neighboring node B is P. A→B The transition probability from the current node A to the neighboring node C is P. A→C The transition probability from the current node A to the neighboring node D is P. A→D Assume P A→B P A→C P A→D China P A→B If the value is the largest, then the adjacent node B is determined as the next path node, that is, the path extends from node A to node B, and the current risk propagation path is A→B.

[0055] Step S204: Determine whether the preset path extension termination condition is met.

[0056] If the preset path extension termination condition is met, the path extension ends and a complete risk propagation path is obtained. If the preset path extension termination condition is not met, the path node determined in step S203 is taken as the new current node, and the "calculate the risk potential energy difference between the current node and each adjacent node" and subsequent steps in step S201 are executed until the path extension termination condition is met.

[0057] For example, the starting node of the path is node A, and the next path node determined with node A as the current node is node B. After determining node B, it is determined whether the preset path extension termination condition is met. If the preset path extension termination condition is not met, node B is taken as the new current node. For the new current node B, the steps S201 "calculate the risk potential energy difference between the current node and each adjacent node" and S202 are executed to obtain the transition probability distribution between the current node B and each adjacent node, and then the next path node is determined accordingly to realize the path extension. Suppose that the next path node determined with node B as the current node is node C, then the current risk propagation path is updated to A→B→C. Then, it is determined whether the preset path extension termination condition is met. If it is not met, node C is taken as the new current node, and the next path node is determined until the preset path extension termination condition is met.

[0058] The path extension termination condition can be that the path length of the current risk propagation path reaches a preset path length threshold. Of course, this embodiment is not limited to this. For example, the path extension termination condition can also be that the risk potential energy values ​​corresponding to adjacent nodes on the current risk propagation path meet a preset change threshold condition.

[0059] Through the above process, a directional risk transmission path can be generated.

[0060] After obtaining the risk propagation path, enterprise risk warning results can be generated based on the risk propagation path. This process is described in some embodiments of this application.

[0061] like Figure 3 As shown, the process of generating enterprise risk warning results based on the risk propagation path may include: Step S301: Based on the risk propagation path, construct node-level risk feature sequences, path-level risk feature sequences, and theme-level risk feature sequences to obtain hierarchical risk features of the risk propagation path.

[0062] In one possible implementation, the process of constructing node-level risk feature sequences, path-level risk feature sequences, and topic-level risk feature sequences based on the risk propagation path, to obtain hierarchical risk features of the risk propagation path, may include: Step S3011: According to the order of each node on the risk propagation path, form a path node sequence of each node on the risk propagation path.

[0063] Based on the order of the nodes in the risk propagation path, the nodes in the risk propagation path are arranged to obtain the path node sequence.

[0064] For example, the risk propagation path is: Enterprise Node A → Related Entity Node B → Enterprise Node C → Related Entity Node D → Enterprise Node E. This risk propagation path indicates that the risk starts from Enterprise Node A, is transmitted to Enterprise Node C via Related Entity Node B, and then continues to propagate to Enterprise Node E via Related Entity Node D. According to the order of each node on this risk propagation path, the nodes on the risk propagation path are arranged to obtain the path node sequence {A, B, C, D, E}.

[0065] Step S3012: Based on the order of each node on the risk propagation path, form a node type sequence from the node types of each node on the risk propagation path.

[0066] Among them, the node type of one node is an enterprise node or an associated entity node.

[0067] For example, the risk propagation path is: Enterprise Node A → Related Entity Node B → Enterprise Node C → Related Entity Node D → Enterprise Node E. Based on the order of nodes A, B, C, D, and E in the risk propagation path, the node types of nodes A, B, C, D, and E are arranged to obtain the node type sequence {Enterprise Node, Related Entity Node, Enterprise Node, Related Entity Node, Enterprise Node}. If E (EnterpriseNode) represents the enterprise node and R (RelatedEntityNode) represents the related entity node, then the node type sequence is {E, R, E, R, E}.

[0068] Step S3013: Based on the path node sequence and node type sequence, construct node-level risk feature sequence, path-level risk feature sequence and topic-level risk feature sequence to obtain hierarchical risk features of the risk propagation path.

[0069] In one possible implementation, the process of constructing node-level risk feature sequences, path-level risk feature sequences, and topic-level risk feature sequences based on path node sequences and node type sequences to obtain hierarchical risk features of the risk propagation path may include: Step a1: Construct a node-level risk feature sequence based on the node risk potential value corresponding to each node in the path node sequence.

[0070] There are multiple ways to construct a node-level risk feature sequence based on the node risk potential value corresponding to each node in the path node sequence. This embodiment provides the following two implementation methods.

[0071] First implementation method: Based on the order of the nodes in the path node sequence, the node risk potential energy values ​​corresponding to each node in the path node sequence are combined to form a node risk potential energy value sequence, which serves as a node-level risk feature sequence.

[0072] This implementation arranges the node risk potential values ​​corresponding to each node in the path node sequence according to the order of each node in the path node sequence, and obtains the node risk potential value sequence. This node risk potential value sequence is then used as the node-level risk feature sequence.

[0073] For example, if the path node sequence is {A, B, C, D, E}, the node risk potential value corresponding to node A is 0.82, the node risk potential value corresponding to node B is 0.76, the node risk potential value corresponding to node C is 0.69, the node risk potential value corresponding to node D is 0.63, and the node risk potential value corresponding to node E is 0.58, then the node-level risk feature sequence is {0.82, 0.76, 0.69, 0.63, 0.58}.

[0074] The second implementation method: Based on the order of the nodes in the path node sequence, the node risk potential values ​​corresponding to each node in the path node sequence are used to form a node-level initial risk feature sequence; based on the order of the nodes in the path node sequence, a position index is assigned to each node in the path node sequence to obtain a position index sequence; the node-level initial risk feature sequence is combined with the position index and node risk potential value of the corresponding node in the position index sequence to obtain the final node-level risk feature sequence.

[0075] This implementation arranges the node risk potential values ​​corresponding to each node in the path node sequence according to the order of the nodes, thus obtaining a node risk potential value sequence. This node risk potential value sequence is used as the node-level initial risk feature sequence. At the same time, according to the order of the nodes in the path node sequence, a position index is assigned to each node in the path node sequence to obtain a position index sequence. After obtaining the node-level initial risk feature sequence and the position index sequence, the node-level initial risk feature sequence and the position index sequence are aligned and fused to obtain the final node-level risk feature sequence.

[0076] For example, the path node sequence is {A, B, C, D, E}. The node risk potential value corresponding to node A is 0.82, that of node B is 0.76, that of node C is 0.69, that of node D is 0.63, and that of node E is 0.58. Then, according to the order of the nodes in the path node sequence, the node risk potential values ​​corresponding to each node are arranged sequentially, resulting in the node risk potential value sequence {0.82, 0.76, 0.69, 0.63, 0.58}. This node risk potential value sequence serves as the initial risk feature sequence at the node level. Simultaneously, based on the path node sequence... The order of nodes is determined by assigning position indices to each node in the path node sequence. For example, node A is assigned position index "1", node B is assigned position index "2", node C is assigned position index "3", node D is assigned position index "4", and node E is assigned position index "5", resulting in the position index sequence {1, 2, 3, 4, 5}. Finally, the initial risk feature sequence at the node level {0.82, 0.76, 0.69, 0.63, 0.58} is combined with the position index and node risk potential value of the corresponding node in the position index sequence {1, 2, 3, 4, 5}. Specifically, the node risk potential value of node A, 0.82, is combined with the position index "1" to obtain (0.82, 1) Combine the node risk potential energy value of 0.76 corresponding to node B with the position index "2" to get (0.76, 2); combine the node risk potential energy value of 0.69 corresponding to node C with the position index "3" to get (0.69, 3); combine the node risk potential energy value of 0.63 corresponding to node D with the position index "4" to get (0.63, 4); combine the node risk potential energy value of 0.58 corresponding to node E with the position index "5" to get (0.58, 5). The final node-level risk feature sequence is {(0.82, 1), (0.76, 2), (0.69, 3), (0.63, 4), (0.58, 5)}.

[0077] Step a2: Traverse the adjacent node type pairs in the node type sequence, form a triplet by combining the adjacent node type pairs and the relationship type between the two nodes corresponding to the adjacent node type pairs. Arrange the obtained triplets according to the order of each node on the risk propagation path to obtain the path-level structural feature sequence.

[0078] For example, given a node type sequence of {E, R, E, R, E}, traversing adjacent node type pairs in this sequence: the first adjacent node type pair encountered is (E, R), and the relationship type between the two nodes (A, B) corresponding to the adjacent node type pair (E, R) is rel_type. A-BThen E, R and rel_type A-B Form a triple (E, R, rel_type) A-B The next adjacent node type pair encountered is (R, E), and the relationship type between the two nodes (B, C) corresponding to the adjacent node type pair (R, E) is rel_type. B-C Then R, E and rel_type B-C Form a triple (R, E, rel_type) B-C The traversal continues until the next adjacent node type pair is (E, R). The relationship type between the two nodes (C, D) corresponding to the adjacent node type pair (E, R) is rel_type. C-D Then E, R and rel_type C-D Form a triple (E, R, rel_type) C-D The last adjacent node type pair encountered is (R, E), and the relationship type between the two nodes (D, E) corresponding to the adjacent node type pair (R, E) is rel_type. D-E Then R, E and rel_type D-E Form a triple (R, E, rel_type) D-E After the traversal is complete, the final sequence of triples is obtained: [(E, R, rel_type)] A-B ), (R, E, rel_type B-C ), (E, R, rel_type C-D ), (R, E, rel_type D-E The triplet sequence is the path-level structural feature sequence.

[0079] Step a3: Align and fuse the node-level risk feature sequence with the path-level structural feature sequence to obtain the path-level risk feature sequence.

[0080] Each risk feature in the path-level risk feature sequence includes adjacent node type pairs, the risk features of the two nodes corresponding to the adjacent node type pairs, and the relationship type between the two nodes corresponding to the adjacent node type pairs.

[0081] For example, the node-level risk feature sequence is {f A , f B , f C , f D , f E}, path-level structural feature sequence [(E, R, rel_type A-B), (R, E, rel_type B-C ), (E, R, rel_type C-D ), (R, E, rel_type D-E The node-level risk feature sequence is aligned and fused with the path-level structural feature sequence to obtain {(E, R, f]}. A , F B rel_type A-B ), (R, E, f B , f C rel_type B-C ), (E, R, f C , F D rel_type C-D ), (R, E, F D , f E rel_type D-E This sequence is the path-level risk feature sequence.

[0082] Step a4: Merge risk features with the same relationship type in the path-level risk feature sequence to obtain the topic-level risk feature sequence.

[0083] For example, the path-level risk feature sequence is {(E, R, f}. A , f B rel_type A-B ), (R, E, f B ,f C rel_type B-C ), (E, R, f C , f D rel_type C-D ), (R, E, f D , f E rel_type D-E )}, assuming rel_type A-B With rel_type D-E If they are the same, then the risk characteristics (E, R, f) will be... A , f B rel_type A-B ) and (R, E, F) D ,f E rel_type D-E ) fusion, combining the fused features (R, E, f) B , f C rel_type B-C) and (E, R, f C , F D rel_type C-D The sequence composed of these elements is used as a topic-level risk feature sequence.

[0084] Through the above steps a1 to a4, node-level risk features, path-level risk feature sequences, and theme-level risk feature sequences can be obtained.

[0085] In one possible implementation, the node-level risk features, path-level risk feature sequences, and theme-level risk feature sequences obtained through steps a1 to a4 above can be used as the final node-level risk features, path-level risk feature sequences, and theme-level risk feature sequences, thereby obtaining the hierarchical risk features of the risk propagation path.

[0086] In one possible implementation, after obtaining the node-level risk feature sequence, path-level risk feature sequence, and theme-level risk feature sequence through steps a1 to a4, the weight vectors corresponding to the node-level risk feature sequence, path-level risk feature sequence, and theme-level risk feature sequence can be determined. Then, the determined weight vectors are used to weight the corresponding risk feature sequences, and the weighted node-level risk feature sequence, path-level risk feature sequence, and theme-level risk feature sequence are used as the final node-level risk feature sequence, path-level risk feature sequence, and theme-level risk feature sequence, thereby obtaining the hierarchical risk features of the risk propagation path.

[0087] When determining the weight vectors corresponding to the node-level risk feature sequence, path-level risk feature sequence, and theme-level risk feature sequence, such as Figure 4 As shown, the node-level risk feature sequence can be input into the node-level attention network to obtain the node-level attention weight vector, which is the weight vector corresponding to the node-level risk feature sequence. The path-level risk feature sequence can be input into the path-level attention network to obtain the path-level attention weight vector, which is the weight vector corresponding to the path-level risk feature sequence. The topic-level risk feature sequence can be input into the topic-level attention network to obtain the topic-level attention weight vector, which is the weight vector corresponding to the topic-level risk feature sequence.

[0088] Step S302: Generate enterprise risk warning results based on the hierarchical risk characteristics of the risk propagation path.

[0089] In one possible implementation, the process of generating enterprise risk warning results based on the hierarchical risk characteristics of the risk propagation path may include: Step S3021: Merge the node-level risk feature sequence, path-level risk feature sequence, and theme-level risk feature sequence in the hierarchical risk features of the risk propagation path to obtain the comprehensive risk features of the risk propagation path.

[0090] Optionally, after obtaining the hierarchical risk characteristics of the risk propagation path, the node-level risk characteristic sequence, path-level risk characteristic sequence, and theme-level risk characteristic sequence in the hierarchical risk characteristics can be spliced ​​together to obtain the comprehensive risk characteristics of the risk propagation path.

[0091] Step S3022: Determine the enterprise risk score based on the comprehensive risk characteristics of the risk propagation path.

[0092] It can perform risk assessment mapping calculations on the comprehensive risk characteristics of risk propagation paths to obtain enterprise risk scores.

[0093] Optionally, the comprehensive risk characteristics of the risk propagation path can be input into the risk assessment and mapping module for numerical mapping calculation to obtain the enterprise risk score.

[0094] When inputting the comprehensive risk characteristics of the risk propagation path into the risk determination mapping module, the comprehensive risk characteristics of the risk propagation path can first be normalized to obtain standard comprehensive risk characteristics. Then, the standard comprehensive risk characteristics can be input into the risk determination mapping module for numerical mapping calculation. Of course, this embodiment is not limited to this, and the comprehensive risk characteristics of the risk propagation path can also be directly input into the risk determination mapping module for numerical mapping calculation.

[0095] Step S3023: Determine the risk scoring interval to which the enterprise's risk score belongs from multiple preset risk scoring intervals.

[0096] Multiple risk scoring intervals can be preset, each corresponding to a risk level. After obtaining the enterprise's risk score, the risk scoring interval to which the enterprise's risk score belongs can be determined from the multiple preset risk scoring intervals.

[0097] Step S3024: Generate enterprise risk warning results that include the enterprise risk score and the risk score range to which the enterprise risk score belongs.

[0098] After obtaining the risk score range to which the enterprise risk score belongs, an enterprise risk warning result can be generated, which includes the enterprise identifier, the enterprise risk score, and the risk score range to which the enterprise risk score belongs.

[0099] In some embodiments of this application, after obtaining the enterprise risk warning result, risk response strategy parameters can be determined based on the risk propagation path and the enterprise risk warning result.

[0100] In one possible implementation, such as Figure 5 As shown, the process of determining risk response strategy parameters based on risk propagation paths and enterprise risk warning results may include: Step S501: Construct the path parameter sequence corresponding to the risk propagation path.

[0101] The path parameter sequence corresponding to the risk propagation path includes the node risk potential value corresponding to each node on the risk propagation path and the relationship weight between adjacent nodes on the risk propagation path.

[0102] For example, the risk propagation path is: Enterprise Node A → Related Entity Node B → Enterprise Node C → Related Entity Node D → Enterprise Node E. Here, the node risk potential value corresponding to Enterprise Node A is 0.82, the node risk potential value corresponding to Related Entity Node B is 0.76, the node risk potential value corresponding to Enterprise Node C is 0.69, the node risk potential value corresponding to Related Entity Node D is 0.63, and the node risk potential value corresponding to Enterprise Node E is 0.58. The relationship weight between Enterprise Node A and Related Entity Node B is 0.91, the relationship weight between Related Entity Node B and Enterprise Node C is 0.87, the relationship weight between Enterprise Node C and Related Entity Node D is 0.83, and the relationship weight between Related Entity Node D and Enterprise Node E is 0.79. Therefore, the path parameter sequence corresponding to this risk propagation path can be {(Node A, 0.82), (Relationship AB, 0.91), (Node B, 0.76), (Relationship BC, 0.87), (Node C, 0.69), (Relationship CD, 0.83), (Node D, 0.83), (Node E, 0.84)}. 0.63), (relation DE, 0.79), (node ​​E, 0.58)}.

[0103] Step S502: Determine the risk response strategy parameters based on the path parameter sequence and the enterprise risk warning results.

[0104] The risk response strategy parameters in this embodiment may include, but are not limited to, some or all of the following parameters: risk handling strategy type parameter, risk response priority parameter, risk control resource allocation parameter, risk propagation blocking parameter, and risk monitoring update parameter.

[0105] Among them, the risk handling strategy type parameter indicates the way the risk is handled, such as risk avoidance, risk mitigation, risk transfer, or risk monitoring. The risk response priority parameter indicates the urgency of handling the current risk event among all risks. The risk control resource allocation parameter is the proportion or quantity of resources allocated to handle the current risk, such as the proportion of monitoring resources, the quantity of review resources, and the quantity of handling tasks. The risk propagation blocking parameter indicates the key node locations that need to be controlled or blocked in the risk propagation path. The risk monitoring update parameter is the update frequency or monitoring cycle of subsequent risk monitoring.

[0106] In one possible implementation, the risk management strategy type parameter, which indicates the risk handling method, can be determined based on the enterprise risk score and risk score range in the enterprise risk warning results.

[0107] For example, in the enterprise risk warning result, the enterprise risk score is 0.82, and the risk score range to which the enterprise risk score belongs is the risk score range corresponding to level 3 risk. Then the risk handling strategy type parameter is determined to be "02", where "02" represents a risk mitigation strategy.

[0108] In one possible implementation, the risk response priority parameter can be calculated based on the enterprise risk score and the length of the risk propagation path in the enterprise risk warning results, according to a preset risk response priority calculation rule. It should be noted that the length of the risk propagation path is the total number of nodes along the risk propagation path.

[0109] For example, if the enterprise risk score is 0.82 and the length of the risk propagation path is 5, and the preset priority calculation rule is: risk response priority = enterprise risk score × path length, then the risk response priority parameter = 0.82 × 5 = 4.10.

[0110] In one possible implementation, risk propagation blocking parameters can be determined based on the node risk potential values ​​in the path parameter sequence. Specifically, key risk propagation nodes can be identified based on the node risk potential values ​​in the path parameter sequence, and these key risk propagation nodes can be used as risk propagation blocking parameters.

[0111] For example, the node risk potential values ​​in the path parameter sequence are: (node ​​A, 0.82), (node ​​B, 0.76), (node ​​C, 0.69), (node ​​D, 0.63), and (node ​​E, 0.58). Based on the node risk potential values ​​in the path parameter sequence, the key risk propagation nodes on the risk propagation path can be determined according to the preset key risk propagation node determination rules. Assuming that the key risk propagation node determination rule is to determine the two nodes with the largest node risk potential values ​​as key risk propagation nodes, then nodes A and B, corresponding to the two largest node risk potential values ​​in the path parameter sequence, are determined as key risk propagation nodes, and the risk propagation blocking parameters are {A, B}.

[0112] In one possible implementation, risk control resource allocation parameters can be determined based on the relational weights in the path parameter sequence.

[0113] For example, the relation weights in the path parameter sequence are: (relation AB, 0.91), (relation BC, 0.87), (relation CD, 0.83), and (relation DE, 0.79). The average of the four relation weights can be calculated, and the average of the four relation weights is (0.91+0.87+0.83+0.79) / 4=0.85. The obtained average is used as the resource allocation intensity coefficient, and then the risk control resource allocation parameters are generated based on the resource allocation intensity coefficient.

[0114] In one possible implementation, risk monitoring update parameters can be determined based on the risk score range in the enterprise risk warning results.

[0115] For example, if the risk score range to which the enterprise risk score belongs in the enterprise risk warning result is the risk score range corresponding to level 3 risk, then the risk monitoring update parameter is determined to be a high-frequency monitoring mode (e.g., data collection is performed once every 1 time unit, and status update is performed once every 2 time units).

[0116] The risk response strategy parameters can be obtained through the above process. Below is an example of risk response strategy parameters determined based on the risk propagation path and the enterprise's risk warning results: Risk management strategy type parameter: Risk mitigation (code 02); Risk response priority parameter: 4.10; Risk transmission blocking parameters: {A, B}; Risk control resource allocation parameters: 85% for monitoring and 80% for disposal. Risk monitoring update parameters: high-frequency monitoring mode.

[0117] The enterprise risk early warning method provided in this invention has the following advantages: Firstly, traditional enterprise risk warnings often focus on the financial indicators or operating conditions of a single enterprise, making it difficult to capture the implicit risk transmission between enterprises through equity, guarantees, related-party transactions, etc. This invention constructs an enterprise risk relationship diagram, placing enterprises within an enterprise relationship network for consideration, which can effectively capture the risk correlation between enterprises.

[0118] Secondly, by introducing node risk potential energy values ​​as guiding signals, each node is given a "potential energy" attribute in the enterprise risk relationship diagram, so that risk propagation is no longer a blind spread, but a directional movement guided by the node risk potential energy values. This mechanism simulates the inherent law of risk transmission from high potential energy to low potential energy in actual networks, and the generated propagation path is more directional and interpretable.

[0119] Third, by performing structural analysis and hierarchical feature mapping on the risk transmission path, the complex risk transmission path is deconstructed into three levels of risk characteristics: "node level - path level - theme level". This hierarchical expression not only retains the individual attributes of micro nodes, but also depicts the transmission mode of meso path, and further refines the semantic information of macro risk theme, thus forming a multi-level representation of risk transmission path.

[0120] Fourth, this invention not only generates enterprise risk warning results, but also generates risk response strategy parameters based on risk propagation paths and enterprise risk warning results. It integrates "risk identification - warning - strategy generation" into an integrated modeling process, which changes the functional limitations of traditional warning schemes that are limited to risk alerts but fail to provide handling guidance. It provides clear operational guidance for subsequent risk handling, thereby reducing the time delay between risk identification and handling response.

[0121] This application also provides an enterprise risk early warning device, such as... Figure 6 As shown, the enterprise risk early warning device may include: an enterprise risk relationship diagram construction module 601, a risk characteristic data acquisition module 602, a risk potential value determination module 603, a risk propagation path generation module 604, and an enterprise risk early warning result generation module 605.

[0122] The Enterprise Risk Relationship Graph Construction Module 601 is used to construct an enterprise risk relationship graph with enterprises and their related entities as nodes and the relationships between nodes as edges.

[0123] The risk characteristic data acquisition module 602 is used to acquire the risk characteristic data corresponding to each node in the enterprise risk relationship diagram.

[0124] The risk potential value determination module 603 is used to determine the node risk potential value corresponding to each node in the enterprise risk relationship diagram based on the risk characteristic data corresponding to each node in the enterprise risk relationship diagram.

[0125] Among them, the node risk potential value represents the degree of risk aggregation or the potential intensity of risk propagation of the corresponding node.

[0126] The risk propagation path generation module 604 is used to determine the starting node of the path based on the node risk potential value corresponding to each node in the enterprise risk relationship diagram, and to extend the path along the relationship between nodes starting from the starting node, using the node risk potential value as the guiding signal for path extension, thereby generating a directional risk propagation path.

[0127] The Enterprise Risk Warning Result Generation Module 605 is used to generate enterprise risk warning results based on the risk propagation path.

[0128] In one possible implementation, such as Figure 6 As shown, the enterprise risk early warning device may also include: a response strategy parameter determination module 606.

[0129] The response strategy parameter determination module 606 is used to determine the risk response strategy parameters based on the risk propagation path and the enterprise risk warning results.

[0130] In one possible implementation, the risk characteristic data acquisition module 602, when acquiring the risk characteristic data corresponding to each node in the enterprise risk relationship diagram, is specifically used for: For each node in the enterprise risk relationship diagram, obtain the multi-source risk characteristic data corresponding to that node; The multi-source risk characteristic data includes some or all of the following characteristic data: enterprise operation characteristic data reflecting the enterprise's operating status at the corresponding node; interaction characteristic data reflecting the interaction between the corresponding node and related nodes; historical risk characteristic data reflecting the historical risk situation of the corresponding node; and structural characteristic data reflecting the position structure of the corresponding node in the enterprise risk relationship diagram.

[0131] In one possible implementation, the risk potential value determination module 603, when determining the node risk potential value corresponding to each node in the enterprise risk relationship diagram based on the risk characteristic data corresponding to each node, is specifically used for: By calling the large model, and based on the risk characteristic data corresponding to each node in the enterprise risk relationship diagram, the node risk potential value corresponding to each node in the enterprise risk relationship diagram is determined.

[0132] In one possible implementation, adjacent nodes in the enterprise risk relationship diagram are configured with relationship weights. The risk propagation path generation module 604, starting from the path's initial node and using the node's risk potential value as a guiding signal for path extension, extends the path along the relationships between nodes to generate a directional risk propagation path. Specifically, it is used for: Take the starting node of the path as the current node, and calculate the risk potential energy difference between the current node and each of its adjacent nodes. Based on the calculated risk potential energy difference and the relationship weights between the current node and each neighboring node, calculate the transition probability distribution from the current node to each neighboring node. Based on the calculated transition probability distribution, the next path node is determined from each of the current node's neighboring nodes to extend the path; If the preset path extension termination condition is not met, the determined next path node will be used as the new current node. The risk potential energy difference between the current node and each adjacent node will be calculated and subsequent steps will be performed until the path extension termination condition is met, and a complete risk propagation path will be obtained.

[0133] In one possible implementation, the enterprise risk warning result generation module 605 includes: a hierarchical risk feature construction module and an enterprise risk warning result generation module.

[0134] The hierarchical risk feature construction module is used to construct node-level risk feature sequences, path-level risk feature sequences, and theme-level risk feature sequences based on the risk propagation path, thereby obtaining hierarchical risk features of the risk propagation path.

[0135] The Enterprise Risk Warning Result Generation Module is used to generate enterprise risk warning results based on the hierarchical risk characteristics of the risk propagation path.

[0136] In one possible implementation, the hierarchical risk feature construction module, when constructing node-level risk feature sequences, path-level risk feature sequences, and topic-level risk feature sequences based on the risk propagation path to obtain hierarchical risk features of the risk propagation path, is specifically used for: Based on the order of the nodes in the risk propagation path, the nodes in the risk propagation path are arranged into a path node sequence. Based on the order of the nodes in the risk propagation path, the node types of each node in the risk propagation path are arranged into a node type sequence. Based on the path node sequence and node type sequence, node-level risk feature sequence, path-level risk feature sequence, and theme-level risk feature sequence are constructed to obtain hierarchical risk features of the risk propagation path.

[0137] In one possible implementation, adjacent nodes in the enterprise risk relationship graph are configured with relationship types. The hierarchical risk feature construction module, when constructing node-level risk feature sequences, path-level risk feature sequences, and topic-level risk feature sequences based on the path node sequence and the node type sequence, is specifically used for: Based on the node risk potential value corresponding to each node in the path node sequence, construct a node-level risk feature sequence; Traverse the adjacent node type pairs in the node type sequence, form a triplet by combining the adjacent node type pairs and the relationship type between the two nodes corresponding to the adjacent node type pairs, and form a path-level structural feature sequence by combining the obtained triplets according to the order of each node on the risk propagation path. The node-level risk feature sequence is aligned and fused with the path-level structural feature sequence to obtain the path-level risk feature sequence. Each risk feature in the path-level risk feature sequence includes a pair of adjacent node types, as well as the relationship type of the two nodes corresponding to the pair of adjacent node types and the node risk feature. Risk features with the same relationship type in the path-level risk feature sequence are merged to obtain the topic-level risk feature sequence.

[0138] In one possible implementation, the hierarchical risk feature construction module, when constructing a node-level risk feature sequence based on the node risk potential value corresponding to each node in the path node sequence, is specifically used for: Based on the order of each node in the path node sequence, the node risk potential values ​​corresponding to each node in the path node sequence are used to form a node-level initial risk feature sequence. Based on the order of the nodes in the path node sequence, a position index is assigned to each node in the path node sequence to obtain a position index sequence. The initial node-level risk feature sequence is combined with the location index and node risk potential value of the corresponding node in the location index sequence to obtain the final node-level risk feature sequence.

[0139] In one possible implementation, the hierarchical risk feature building module is also used for: Inputting the node-level risk feature sequence into the node-level attention network yields the node-level attention weight vector; inputting the path-level risk feature sequence into the path-level attention network yields the path-level attention weight vector; inputting the topic-level risk feature sequence into the topic-level attention network yields the topic-level attention weight vector. The node-level risk feature sequence is weighted using a node-level attention weight vector; the path-level risk feature sequence is weighted using a path-level attention weight vector; and the topic-level risk feature sequence is weighted using a topic-level attention weight vector. The weighted node-level risk feature sequence is used as the final node-level risk feature sequence, the weighted path-level risk feature sequence is used as the final path-level risk feature sequence, and the weighted theme-level risk feature sequence is used as the final theme-level risk feature sequence.

[0140] In one possible implementation, when generating enterprise risk warning results based on the hierarchical risk characteristics of the risk propagation path, the enterprise risk warning result generation module is specifically used for: By fusing the node-level risk feature sequence, path-level risk feature sequence, and theme-level risk feature sequence from the hierarchical risk characteristics of the risk propagation path, a comprehensive risk characteristic of the risk propagation path is obtained. Based on the comprehensive risk characteristics of the risk propagation path, determine the enterprise risk score; Determine the risk scoring range to which the enterprise's risk score belongs from a set of preset risk scoring ranges; Generate enterprise risk warning results that include the enterprise risk score and the risk score range to which the enterprise risk score belongs.

[0141] In one possible implementation, the response strategy parameter determination module 606, when determining risk response strategy parameters based on the risk propagation path and the enterprise risk warning results, is specifically used for: Construct a path parameter sequence corresponding to the risk propagation path. The path parameter sequence includes the node risk potential value corresponding to each node on the risk propagation path and the relationship weight between adjacent nodes on the risk propagation path. Based on the path parameter sequence and enterprise risk warning results, determine the parameters for risk response strategies.

[0142] In one possible implementation, the response strategy parameter determination module 606, when determining the risk response strategy parameters based on the path parameter sequence and the enterprise risk warning results, is specifically used for: Based on the enterprise risk score and risk score range in the enterprise risk warning results, determine the risk handling strategy type parameter used to indicate the risk handling method; And / or, determine the risk response priority parameters based on the path length of the risk propagation path and the enterprise risk score in the enterprise risk warning results; And / or, determine the risk propagation blocking parameters based on the node risk potential values ​​in the path parameter sequence; And / or, determine the risk control resource allocation parameters based on the relation weights in the path parameter sequence; And / or, determine the risk monitoring update parameters based on the risk score range in the enterprise risk warning results.

[0143] The enterprise risk early warning device provided in this application places enterprises within a relational network, effectively capturing risk correlations between enterprises. By generating risk propagation paths guided by node risk potential values, it can reveal the direction and trajectory of risk transmission from the source outwards. Based on the risk propagation path, it can issue early warnings before risks spread to target enterprises, realizing a shift from post-event handling to pre-event prevention, significantly improving the timeliness and foresight of early warnings. Based on the risk propagation path and enterprise risk early warning results, it generates risk response strategy parameters, providing clear operational guidance for subsequent risk handling.

[0144] This application also provides an electronic device, which may include at least one processor and a memory connected to the processor.

[0145] The processor may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application; the memory may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage.

[0146] The memory is used to store computer programs, and the processor is used to execute the computer programs so that the electronic device can implement the enterprise risk warning method provided in the above embodiments.

[0147] This application also provides a computer storage medium that carries one or more computer programs. When one or more computer programs are executed by an electronic device, the electronic device can implement the enterprise risk warning method provided in the above embodiments.

[0148] This application also provides a computer program product, including computer-readable instructions, which, when executed on an electronic device, enable the electronic device to implement the enterprise risk warning method provided in the above embodiments.

[0149] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0150] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0151] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0152] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

Claims

1. A method for enterprise risk early warning, characterized in that, include: Using enterprises and their associated entities as nodes, and the relationships between nodes as edges, an enterprise risk relationship graph is constructed, wherein adjacent nodes in the enterprise risk relationship graph are configured with relationship weights; Obtain the risk characteristic data corresponding to each node in the enterprise risk relationship diagram; The large model is invoked, and the risk potential value of each node in the enterprise risk relationship diagram is determined based on the risk characteristic data corresponding to each node. The node risk potential value represents the degree of risk aggregation or the potential intensity of risk propagation of the corresponding node. The starting node of the path is determined based on the node risk potential value corresponding to each node in the enterprise risk relationship diagram. Starting from the starting node, the node risk potential value is used as the guiding signal for path extension. The path is extended along the relationship between nodes to generate a directional risk propagation path. Based on the described risk propagation path, generate enterprise risk warning results; The step of starting from the path's initial node, using the node's risk potential value as a guiding signal for path extension, and extending the path along the relationships between nodes to generate a directional risk propagation path includes: Taking the starting node of the path as the current node, calculate the risk potential energy difference between the current node and each adjacent node; based on the calculated risk potential energy difference and the relationship weight between the current node and each adjacent node, calculate the transition probability distribution from the current node to each adjacent node. Based on the calculated transition probability distribution, the next path node is determined from each of the adjacent nodes of the current node to extend the path; if the preset path extension termination condition is not met, the next path node is taken as the new current node, and the calculation of the risk potential energy difference between the current node and each of the adjacent nodes and subsequent steps are performed until the path extension termination condition is met, and a complete risk propagation path is obtained.

2. The enterprise risk early warning method according to claim 1, characterized in that, Also includes: Based on the risk propagation path and the enterprise risk warning results, determine the parameters of the risk response strategy; The step of determining risk response strategy parameters based on the risk propagation path and the enterprise risk warning results includes: Construct a path parameter sequence corresponding to the risk propagation path. The path parameter sequence includes the node risk potential value corresponding to each node on the risk propagation path and the relationship weight between adjacent nodes on the risk propagation path. Based on the path parameter sequence and the enterprise risk warning results, determine the risk response strategy parameters; The risk response strategy parameters include some or all of the following parameters: risk handling strategy type parameter, risk response priority parameter, risk control resource allocation parameter, risk propagation blocking parameter, and risk monitoring update parameter; the risk handling strategy type parameter is a parameter indicating the risk handling method, the risk response priority parameter is a parameter indicating the urgency of risk handling, the risk control resource allocation parameter is a resource allocation ratio or resource quantity, and the risk propagation blocking parameter is a node in the risk propagation path that needs to be controlled or blocked.

3. The enterprise risk early warning method according to claim 1, characterized in that, The step of obtaining the risk characteristic data corresponding to each node in the enterprise risk relationship diagram includes: For each node in the enterprise risk relationship diagram, obtain the multi-source risk feature data corresponding to the node; The multi-source risk characteristic data includes some or all of the following characteristic data: enterprise operation characteristic data reflecting the enterprise operation status of the corresponding node, interaction characteristic data reflecting the interaction between the corresponding node and related nodes, historical risk characteristic data reflecting the historical risk situation of the corresponding node, and structural characteristic data reflecting the position structure of the corresponding node in the enterprise risk relationship diagram.

4. The enterprise risk early warning method according to claim 1, characterized in that, The step of generating enterprise risk warning results based on the risk propagation path includes: Based on the risk propagation path, node-level risk feature sequences, path-level risk feature sequences, and theme-level risk feature sequences are constructed to obtain the hierarchical risk features of the risk propagation path. Based on the hierarchical risk characteristics of the risk propagation path, enterprise risk warning results are generated.

5. The enterprise risk early warning method according to claim 4, characterized in that, The process involves constructing node-level risk feature sequences, path-level risk feature sequences, and topic-level risk feature sequences based on the risk propagation path, resulting in hierarchical risk features of the risk propagation path, including: Based on the order of the nodes on the risk propagation path, the nodes on the risk propagation path are arranged into a path node sequence. Based on the order of the nodes on the risk propagation path, the node types of the nodes on the risk propagation path are arranged into a node type sequence. Based on the path node sequence and the node type sequence, a node-level risk feature sequence, a path-level risk feature sequence, and a topic-level risk feature sequence are constructed to obtain the hierarchical risk features of the risk propagation path.

6. The enterprise risk early warning method according to claim 5, characterized in that, The adjacent nodes in the enterprise risk relationship diagram are configured with relationship types; The step of constructing node-level risk feature sequences, path-level risk feature sequences, and theme-level risk feature sequences based on the path node sequence and the node type sequence includes: Based on the node risk potential value corresponding to each node in the path node sequence, a node-level risk feature sequence is constructed. Traverse the adjacent node type pairs in the node type sequence, form a triplet by combining the adjacent node type pairs and the relationship type between the two nodes corresponding to the adjacent node type pairs, and form a path-level structural feature sequence by combining the obtained triplets according to the order of each node on the risk propagation path. The node-level risk feature sequence is aligned and fused with the path-level structural feature sequence to obtain a path-level risk feature sequence. Each risk feature in the path-level risk feature sequence includes a pair of adjacent node types, as well as the relationship type and node risk feature of the two nodes corresponding to the pair of adjacent node types. Risk features with the same relationship type in the path-level risk feature sequence are merged to obtain the topic-level risk feature sequence.

7. The enterprise risk early warning method according to claim 6, characterized in that, The step of constructing a node-level risk feature sequence based on the node risk potential value corresponding to each node in the path node sequence includes: Based on the order of the nodes in the path node sequence, the node risk potential values ​​corresponding to each node in the path node sequence are used to form a node-level initial risk feature sequence. Based on the order of the nodes in the path node sequence, a position index is assigned to each node in the path node sequence to obtain a position index sequence; The node-level initial risk feature sequence is combined with the location index and node risk potential value of the corresponding node in the location index sequence to obtain the final node-level risk feature sequence.

8. The enterprise risk early warning method according to claim 6, characterized in that, The step of constructing node-level risk feature sequences, path-level risk feature sequences, and theme-level risk feature sequences based on the path node sequence and the node type sequence further includes: The node-level risk feature sequence is input into a node-level attention network to obtain a node-level attention weight vector; the path-level risk feature sequence is input into a path-level attention network to obtain a path-level attention weight vector; the topic-level risk feature sequence is input into a topic-level attention network to obtain a topic-level attention weight vector. The node-level risk feature sequence is weighted using the node-level attention weight vector; the path-level risk feature sequence is weighted using the path-level attention weight vector; the topic-level risk feature sequence is weighted using the topic-level attention weight vector. The weighted node-level risk feature sequence is used as the final node-level risk feature sequence, the weighted path-level risk feature sequence is used as the final path-level risk feature sequence, and the weighted theme-level risk feature sequence is used as the final theme-level risk feature sequence.

9. The enterprise risk early warning method according to claim 4, characterized in that, The step of generating enterprise risk warning results based on the hierarchical risk characteristics of the risk propagation path includes: By fusing the node-level risk feature sequence, path-level risk feature sequence, and theme-level risk feature sequence in the hierarchical risk features of the risk propagation path, a comprehensive risk feature of the risk propagation path is obtained. Based on the comprehensive risk characteristics of the aforementioned risk propagation path, determine the enterprise risk score; From a set of multiple preset risk scoring intervals, determine the risk scoring interval to which the enterprise's risk score belongs; Generate enterprise risk warning results that include the enterprise risk score and the risk score range to which the enterprise risk score belongs.

10. The enterprise risk early warning method according to claim 2, characterized in that, The step of determining risk response strategy parameters based on the path parameter sequence and the enterprise risk warning result includes: Based on the enterprise risk score and risk score range in the enterprise risk warning results, determine the risk handling strategy type parameter used to indicate the risk handling method; And / or, based on the path length of the risk propagation path and the enterprise risk score in the enterprise risk warning result, determine the risk response priority parameter; And / or, determine the risk propagation blocking parameters based on the node risk potential values ​​in the path parameter sequence; And / or, determine the risk control resource allocation parameters based on the relation weights in the path parameter sequence; And / or, determine the risk monitoring update parameters based on the risk score range in the enterprise risk warning results.

11. An electronic device, characterized in that, It includes at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is used to execute the computer program to enable the electronic device to implement the enterprise risk warning method as described in any one of claims 1 to 10.

12. A computer storage medium, characterized in that, The storage medium carries one or more computer programs, which, when executed by an electronic device, enable the electronic device to implement the enterprise risk warning method as described in any one of claims 1 to 10.

13. A computer program product, characterized in that, It includes computer-readable instructions that, when executed on an electronic device, cause the electronic device to implement the enterprise risk warning method as described in any one of claims 1 to 10.

Images